ElogDoct: A Tool for Lipophilicity Determination in Drug Discovery. 2. Basicand Neutral Compounds

Franco Lombardo,*,# Marina Y. Shalaeva,# Karl A. Tupper,# and Feng Gao‡

Molecular Properties Group and Mathematical and Statistical Sciences Group, Pfizer Global Research and Development,Groton Laboratories, Groton, Connecticut 06340

Received March 5, 2001

We present an RP-HPLC method for the determination of the octanol-water distributioncoefficients at pH 7.4, as log Doct

7.4 values, for neutral and basic drugs, which combines ease ofoperation with high accuracy. The method is shown to work for a training set of 90 moleculescomprised largely of drugs, and it was also applied to a test set of 10 proprietary compounds.This work expands the applicability of the method presented in our earlier report, for thedetermination of logPoct for neutral compounds (J. Med. Chem. 2000, 43, 2922-2928), and itoffers the same general features but widens the scope. Generally, the method (i) is compoundsparing (e1 mL of a 50-100 µg/mL solution needed), (ii) is insensitive to concentration andphase ratio effects observed in some shake-flask determinations, (iii) is amenable to rapiddeterminations (e20 min on average), (iv) is insensitive to impurities, (v) possesses a widelipophilicity range (>7 log Doct

7.4 units), and (vi) offers a good accuracy, (vii) an excellentreproducibility, (viii) and an excellent potential for automation. To the best of our knowledge,a similar performance, on a set of noncongeneric drugs, has not been previously reported. Werefer to the value generated via this method as ElogDoct.

IntroductionThe importance of lipophilicity can be understood, for

example, by considering the correlation between highlipophilicity and poor solubility, which has generallybeen explored with neutral solutes.1 Furthermore, li-pophilicity has been shown to be of paramount impor-tance in several other ADME aspects, that is, absorp-tion, distribution, metabolism, and excretion. It isgenerally held that very lipophilic compounds are“preferred” targets for metabolism, often leading to highclearance values and, quite often, lipophilicity positivelycorrelates with a high plasma protein binding.2-5 Alarge volume of distribution, probably due to a highfraction of the compound bound to tissues, is oftenobserved for lipophilic compounds.4 At physiological pHmany basic or acidic drugs are ionized, and the partitioncoefficient is indeed a distribution coefficient, D, whichis generally taken to be the distribution between anaqueous buffer at pH 7.4 and n-octanol, and it isindicated by the notation Doct

7.4.Scherrer3 defines the distribution coefficient (in the

form of its logarithm) for monoprotic bases as

For monoprotic acids the equation has the same form,except that the exponent is written as pH-pKa. Forpolyprotic compounds the equations become more com-plicated, and these aspects have been described in detailby Avdeef.6 These equations assume that only un-

ionized species partition in the organic phase (octanol).In fact, this may or may not be always true. Dependingon the nature and concentration of the species present(including counterions), there may be other phenomenaat work, such as ion partitioning.

Given the widespread use and application of distribu-tion coefficients, a method that can accurately andrapidly yield log Doct

7.4 values would be a welcome addi-tion to the experimental tools available for physico-chemical properties screening in the discovery setting.We will use, in the rest of this work, the generalnotation logDoct , to mean the values determined at pH7.4.

Although computational packages for the estimationof logDoct are available,7 and this value could also becalculated from estimated logPoct and pKa values, wefind that, for drug molecules, computed values are ofteninaccurate. Depending on the software used, they maydiffer by as much as two to three logDoct units amongdifferent software packages and/or from experimentalvalues, since the accuracy of pKa as well as logPoct hasto be factored in. These methods are, of course, valuablewhen virtual libraries (or individual virtual molecules)are being designed and, with proper training, moreaccurate values might be obtained. However, as earlyas possible and especially if a compound-sparing methodis available, the computed values should be replaced bymeasured values, with particular regard to cases whereintramolecular H-bonding is possible, and/or in thepresence of conformational flexibility, and/or with mol-ecules which can tautomerize. These occurrences typi-cally offer an even greater challenge to fragment-basedsoftware packages. SAR and SPR analyses and alertssuch as the Lipinski “rule of 5”8 would greatly benefitby the introduction of accurate experimental values.

* Address correspondence to Franco Lombardo, PGRD GrotonLaboratories, MS 8118W-220, Groton, CT 06340. Phone: (860) 441-6982. Fax: (860) 715-3345. E-mail: franco•lombardo@groton.pfizer.com.

# Molecular Properties Group.‡ Mathematical and Statistical Sciences Group.

log Doct ) log Poct + log [1/(1 + 10pKa-pH)]

2490 J. Med. Chem. 2001, 44, 2490-2497

10.1021/jm0100990 CCC: $20.00 © 2001 American Chemical SocietyPublished on Web 06/26/2001

The classical shake-flask method, or variations of thismethod which have been described,9 are neither ruggednor rapid enough for medium to high-throughput ap-plications, they are generally more sensitive to impuri-ties and less amenable to automation than are RP-HPLC methods, and they do not usually offer a widedynamic range.

RP-HPLC retention data, expressed either as log k′or k′w, have been shown to correlate well with absoluteand relative lipophilicity values,10 but they have alsobeen criticized as not being a true “replacement” forshake-flask values.9 Part of the criticism may stem fromthe limited scope of some reports, focusing either onfairly simple monofunctional solutes11,12 or classes ofanalogues13 with a limited logPoct range, and in manycases there was only mention of neutral solutes, underthe pH conditions of the method. When these correla-tions were extended beyond classes of analogues, lessencouraging results were obtained.10 Furthermore, inseveral cases, the slope of these correlations was quitedifferent from unity, casting doubts about the differentbalance of forces responsible for the two values. Indeed,LFER analyses have shown that log k′ on typical RP-HPLC systems do not encode the same blend of factorsas does logPoct.14 In particular, log k′ values respond tosolute hydrogen-bond acidity, but logPoct values do not.

k′ represents the capacity factor of the solute at agiven organic solvent concentration, and k′w is thecapacity factor extrapolated to a 0% concentration of theorganic solvent. The observations reported above pointedus toward using the extrapolated value rather than, asreported by Yamagami and co-workers,15 a log k′ value,which is likely to be limited in its applicability to a widevariety of drug-like compounds (amphiprotic) and,therefore, of limited use for the goals we set. As wereported previously,16 a judicious choice of columns andmobile phase, as well as flow rates, would greatlyenhance the performance of RP-HPLC methods.

Another factor of great importance and concern, if thedata were to be used for software training purposes orfor the creation of a large database, is the reproducibilityfrom column to column, which depends on the qualityof the packing chemistry and manufacturing. We haveshown, in the case of the ElogP data,16 that the latteraspect does not appear to be a problem. However, it isadvisable to monitor the column performance for pos-sible deterioration, especially if high-throughput screen-ing is the goal.

The speed of the determination and the ability tohandle diverse structures and lipophilicity values are,of course, of paramount importance in an industrialresearch setting. These aspects translate into the ca-pability of screening, with modest resources, a largenumber of compounds, with a good degree of accuracyacross a wide range of lipophilicity values and hydrogen-bonding properties.

Considering the points discussed above, we set outto develop a method that would be accurate, rapid, andpossess a good dynamic range, together with beingapplicable to a variety of drug-like molecules, and whichwould be robust with respect to ion-pairing and concen-tration-related variability, often observed across differ-ent protocols and laboratories. We present and discussour results in the following section.

Results and Discussion

Minick et al.12,17 had used an MC-8 column, coupledwith the addition of 1-octanol to both components of theeluent, together with the addition of small amounts(0.5% v/v) of n-decylamine in the aqueous phase, andthey obtained a very good correlation but on a limiteddata set and using monofunctional solutes. Previously,Unger18 had obtained seemingly good results with theaddition of a tertiary amine but the data set, albeitmostly comprised of drugs, was not very extensive, interms of structural diversity and logDoct values, withthe latter spanning approximately two units. Further-more, the optimized conditions reported (RP-18) yieldedbetter results with a stationary phase slurry packed inthe presence of octanol, and this procedure still yieldeda fairly high (1.24) intercept, although a slope very closeto unity was obtained. Obviously, there are moreadvanced choices today, in terms of pre-packed RPphases. An embedded polar amide column, such as theLC-ABZ,19 had raised our interest from the work ofPagliara et al.11 and had performed very well for neutraldrug-like solutes in the absence of decylamine.16 Thus,we started our work with an attempt at extendingidentical conditions to ionized solutes. However, despiteuse of several different masking agents, some of whichwere zwitterionic or anionic, we could not achieve highaccuracy in reproducing the logDoct values for setsincluding acidic, neutral, and basic compounds. It hasbeen reported that anions seem prone to specific inter-actions with this stationary phase,11 and our observa-tions would seem to confirm that, especially after afairly extensive screen of potential masking agents. Wehave tried other stationary phases but without anysignificant improvement, and we are of the opinion thatit might be very difficult, by a single method, todetermine the lipophilicity of all three classes of com-pounds across a wide range of structures and lipophi-licity with the high accuracy we had set as a goal.However, neutral and basic compounds seem to com-prise most of the drug-like compounds, and a survey on2500 compounds submitted to our laboratory for variousphysicochemical determinations showed the (strongly)acidic compounds to be a fairly small fraction of about5%. Furthermore, for therapeutic areas such as CNS,basic and neutral compounds generally represent aneven higher percentage of compounds.

We soon found that for a subset of the compoundsreported here, including basic and neutral compounds,n-decylamine would yield improved correlations, al-though it was obviously not necessary for the neutralsolutes.16 To maximize the speed of analysis, while stillretaining a good accuracy, the same flow rate previouslyreported16 was chosen for each range of lipophilicity (seeExperimental Section). We also studied the effect of afurther increase of flow rate on accuracy and performedstatistical analysis in order to determine experimentallipophilicity “thresholds” for each range of conditions,which would yield accurate results. The current “thresh-olds” for lipophilicity ranges are reported in the Experi-mental Section.

Extensive work was conducted using the 90 solutesreported in Table 1, and each log k′w value is theaverage of at least three determinations, on differentcolumns, with an average standard deviation of 0.07.

RP-HPLC Lipophilicity Determination Journal of Medicinal Chemistry, 2001, Vol. 44, No. 15 2491

Table 1. Extrapolated Capacity Factors and logDoct Data for the 90 Solutes Used

compda CAS no. log k′wb sdc ElogDoctd logDoct

e res.f refs

acebutolol 37517-30-9 -0.53 0.04 -0.39 -0.29 0.10 25, 26acetaminophen* 103-90-2 0.15 0.01 0.38 0.51 0.13 27acetophenone 98-86-2 1.21 0.03 1.57 1.58 0.01 28allopurinol* 315-30-0 -0.27 0.01 -0.10 -0.44 -0.34 g, 29alprazolam 28981-97-7 1.73 0.04 2.16 2.12 -0.04 30alprenolol 13655-52-2 0.37 0.04 0.62 0.97 0.35 j, 25amiodarone* 1951-25-3 5.10 0.21 5.95 6.10 0.15 31amlodipine 88150-42-9 1.72 0.05 2.15 1.66 -0.49 g, h, 2, 32antipyrine 60-80-0 0.12 0.03 0.34 0.38 0.04 33atropine 51-55-8 -0.33 0.19 -0.16 -0.55 -0.39 j, 34bifonazole 60628-96-8 4.33 0.07 5.09 4.77 -0.32 35bromazepam 1812-30-2 1.04 0.04 1.38 1.65 0.27 36, 373-bromoquinoline* 5332-24-1 2.53 0.06 3.06 3.03 -0.03 38caffeine 58-08-2 -0.19 0.02 -0.01 -0.07 -0.06 35carbamazepine 298-46-4 1.40 0.03 1.78 2.19 0.41 39chloramphenicol* 56-75-7 1.19 0.07 1.55 1.14 -0.41 353-chlorophenol 108-43-0 2.58 0.01 3.11 2.50 -0.61 28chlorpheniramine 132-22-9 1.20 0.16 1.56 1.41 -0.15 j, 18, 40chlorpromazine 50-53-3 2.66 0.08 3.20 3.38 0.18 18, 26, 41chlorthalidone 77-36-1 0.76 0.10 1.06 1.11 0.05 g, j, 42cimetidine 51481-61-9 0.17 0.03 0.40 0.35 -0.05 jclonidine 4205-90-7 0.07 0.03 0.29 0.62 0.33 43clotrimazole* 23593-75-1 4.03 0.06 4.75 5.20 0.45 16clozapine 5786-21-0 2.82 0.03 3.38 3.13 -0.25 j, 18cocaine 50-36-2 0.24 0.05 0.48 1.05 0.57 44codeine 76-57-3 0.16 0.13 0.39 0.23 -0.16 41, 45cyclothiazide 2259-96-3 2.15 0.09 2.63 2.09 -0.54 jdeprenyl 2323-36-6 2.19 0.11 2.67 2.70 0.03 31desipramine 50-47-5 0.97 0.06 1.30 1.28 -0.02 g, j, 46dexamethasone 50-02-2 1.62 0.04 2.03 1.83 -0.20 47diazepam 439-14-5 2.46 0.03 2.98 2.79 -0.19 483,5-dichlorophenol 591-35-5 3.44 0.08 4.08 3.68 -0.40 38diethylstilbestrol 56-53-1 4.12 0.10 4.85 5.07 0.22 35diltiazem 33286-22-5 1.59 0.22 2.00 2.06 0.06 jdiphenhydramine 58-73-1 1.04 0.06 1.38 1.29 -0.09 g, j, 40disopyramide 3737-09-5 -1.21 0.10 -1.16 -0.66 0.50 jestradiol 50-28-2 3.28 0.08 3.90 4.01 0.11 35fentanyl citrate 990-73-8 1.94 0.09 2.39 2.91 0.52 49flecainide 54143-55-4 0.25 0.06 0.49 0.97 0.48 g, 50fluconazole* 86386-73-4 0.40 0.15 0.66 0.50 -0.16 51griseofulvin 126-07-8 1.73 0.06 2.16 2.18 0.02 35haloperidol 52-86-8 2.00 0.05 2.46 2.98 0.52 g, jhydrocortisone 50-23-7 1.14 0.06 1.49 1.55 0.06 49hydrocortisone-21 acetate 50-03-3 1.64 0.01 2.06 2.19 0.13 49imipramine 50-49-7 1.56 0.19 1.97 2.40 0.43 g, j, 18, 40lidocaine 137-58-6 0.96 0.13 1.29 1.71 0.42 g, 18, 41loratadine 79794-75-5 4.02 0.07 4.74 4.40 -0.34 40lorazepam 846-49-1 2.30 0.03 2.80 2.51 -0.29 52lormetazepam 848-75-9 2.27 0.04 2.77 2.72 -0.05 53methotrimeprazine 60-99-1 2.10 0.27 2.57 2.77 0.20 jmethylthioinosine 342-69-8 0.25 0.02 0.49 0.09 -0.40 16metoclopramide 364-62-5 0.46 0.16 0.73 0.64 -0.09 jmetoprolol 56392-17-7 -0.73 0.08 -0.62 -0.16 0.46 25, 31, 34, 46metronidazole 443-48-1 -0.08 0.02 0.12 -0.02 -0.14 54mexiletine 31828-71-4 0.02 0.03 0.23 0.47 0.24 jmorphine sulfate 64-31-3 0.10 0.10 0.32 0.03 -0.29 41, 44, 45naphthalene 91-20-3 3.03 0.06 3.62 3.37 -0.25 38nicotine 54-11-5 0.02 0.06 0.23 0.40 0.17 g, jnifedipine 21829-25-4 2.34 0.06 2.84 3.17 0.33 16nifuroxime* 6236-05-1 1.14 0.07 1.49 1.28 -0.21 16nitrofurazone 59-87-0 0.29 0.01 0.53 0.23 -0.30 55nizatidine 76963-41-2 -0.13 0.04 0.06 -0.52 -0.58 56omeprazole 73590-58-6 1.59 0.04 2.00 2.30 0.30 g, 57pentoxifylline 6493-05-6 0.03 0.02 0.24 0.29 0.05 58pirenzepine 28797-61-7 -0.15 0.06 0.04 -0.61 -0.65 gprednisolone 50-24-8 1.24 0.05 1.60 1.60 0.00 16prednisone 53-03-2 0.90 0.06 1.22 1.46 0.24 35procainamide 51-06-9 -0.69 0.24 -0.57 -0.91 -0.34 g, 59propafenone 54063-53-5 1.14 0.10 1.49 1.81 0.32 g, jpropranolol 525-66-6 0.64 0.01 0.93 1.26 0.33 g, 25, 26, 31, 34, 46, 60quinidine 56-54-2 1.16 0.12 1.51 2.04 0.53 18, 26, 61quinoline 91-22-5 1.52 0.04 1.92 2.03 0.11 62ranitidine 66357-35-5 -0.63 0.01 -0.50 -0.29 0.21 jrisperidone 106266-06-2 1.23 0.13 1.59 2.04 0.45 gsotalol 3930-20-9 -1.47 0.10 -1.45 -1.35 0.10 25, 31sumatriptan 103628-46-2 -0.54 0.02 -0.40 -1.00 -0.60 g, h, 63

2492 Journal of Medicinal Chemistry, 2001, Vol. 44, No. 15 Lombardo et al.

The table also reports the standard deviation of log k′wfor each compound (N ) 3 to 5), and no large deviationswere observed, regardless of structure and range oflipophilicity, with methotrimeprazine yielding the larg-est standard deviation (sd) with a value of 0.27. It isworth noting that routine potentiometric lipophilicitydeterminations have a typical sd of 0.4 logDoct units, forreplicate determinations, and often they have to beextrapolated to 0% organic solvent from mixed solvents,due to the poor aqueous solubility of many compounds.The fit of the log k′w to the averaged logDoct values isreported below (eq 1), and shown graphically in Figure1.

The slope obtained is very close to unity, with a smallintercept, and these parameters offer a good comparisonof the balance of forces which play a role in the shake-flask vs RP-HPLC distribution coefficient. The questionof the diagnostic importance of the slope has beenstressed by Minick et al.17 Pointing to the work ofMelander et al.,20 these authors state that “...equationscorrelating log k′w and logPoct data represent linear free

energy relationships in which the slope is an estimateof how closely the free energies of the processes com-pare.” A slope close to unity implies that the twoprocesses are homoenergetic, i.e., the free energy changesare the same. Furthermore, if the goal is the determi-nation of the “classical” logDoct, then a slope significantlydifferent from 1 would enhance any error in the deter-mination of logDoct. A slope significantly different fromunity is an indication of a fairly large over- or under-estimation of lipophilicity by the method. Obviously, ifa different scale of lipophilicity is the goal, log k′wvalues could be used as such, or different indices couldbe developed. Valko et al.21 described a chromatographichydrophobicity index (CHI), obtained via a gradient run.In this case a correlation with a “classical” shake flasklogPoct was not necessarily sought, and a self-consistentCHI scale was established. However, logDoct data areso widely used in many correlations by the medicinalchemistry community that a “classical” logDoct value islikely to be desired. To the best of our knowledge noother method capable of encompassing all the accuracyand ruggedness requirements we set as goals for thiswork, including a very practical set of conditions andspeed, has been reported in the literature to date. Byanalogy with our previous work16 we termed the valuesobtained via eq 1 as ElogDoct, and we will refer to themas such for the rest of the discussion.

A plot of residuals vs logDoct values, as in Figure 2,shows that the error distribution is very consistentacross the entire range, and no curvature (larger error)is observed at extreme values. This is important becauseit shows that similarly accurate determinations can beobtained across a dynamic range of 7 logDoct units, andthe standard error is fairly small, considering also thevariability of some of the logDoct data reported in theliterature.

The question might be asked, about whether decyl-amine acts as a modifier other than a masking agentfor potentially ionized silanols, since its absence isdetrimental to the performance of the method. A com-parison between log k′w values obtained under theconditions reported in our previous work in 36 neutralsolutes,16 termed here log k′w(P), and the values ob-tained under the present condition for the same solutes,i.e., log k′w(D), shows that the balance of forces is thesame, as demonstrated by a very small intercept (non-significant) and a slope very close to unity, in eq 2. Since

Table 1. (Continued)

compda CAS no. log k′wb sdc ElogDoctd logDoct

e res.f refs

terbutaline sulfate 23031-32-5 -1.51 0.06 -1.49 -1.35 0.14 31, 46, 64testosterone 58-22-0 2.63 0.04 3.17 3.29 0.12 47tetracaine 94-24-6 1.70 0.07 2.12 2.29 0.17 31thiamphenicol 15318-45-3 -0.05 0.01 0.15 -0.27 -0.42 65thioridazine 50-52-2 2.88 0.10 3.45 3.34 -0.11 j, 66tiapride 51012-32-9 -0.58 0.05 -0.45 -0.90 -0.45 jtiotidine 69014-14-8 0.57 0.01 0.85 0.57 -0.28 gtolnaftate* 2398-96-1 4.53 0.10 5.31 5.40 0.09 16trazodone 19794-93-5 2.45 0.06 2.97 2.54 -0.43 jtriamterene 396-01-0 0.71 0.05 1.01 1.21 0.20 g, j, 42, 67trichlormethiazide 133-67-5 0.26 0.02 0.50 0.43 -0.07 g, jtriflupromazine* 146-54-3 3.05 0.13 3.64 3.61 -0.03 j, 66trimethoprim 738-70-5 0.36 0.02 0.61 0.63 0.02 jzaltidine 85604-00-8 0.53 0.02 0.80 0.74 -0.06 g

a Asterisk denotes standard compound. b Average of three to five determinations c Standard deviation of 3 to 5 log k′w determinations.d Equation 1. e Average logDoct from all the references or methods indicated. f logDoct - ElogDoct. g Shake-vial procedure A. h Shake-vialprocedure B. j Potentiometric determination.

Figure 1. Correlation between logDoct and log k′w for 90solutes.

logDoct ) 1.1267 (( 0.0233)log k′w +0.2075 (( 0.0430) (1)

N ) 90, R2 ) 0.964, R ) 0.982, s ) 0.309,F ) 2339, q2 ) 0.962

RP-HPLC Lipophilicity Determination Journal of Medicinal Chemistry, 2001, Vol. 44, No. 15 2493

we have demonstrated that log k′w under our previousconditions encodes the same balance of forces as in abiphasic octanol-water system,16 we conclude that thevalues obtained by this method are “true” logDoct values.

We have also tested a set of 10 proprietary com-pounds, basic and neutral, with molecular weightsbetween 209 and 532 Da. These compounds are struc-turally dissimilar from the compounds in the trainingset and possess a wide range of functional groups. Theresults of the ElogD determinations are compared, inTable 2, with shake-vial and/or potentiometric deter-minations, and they show a good performance of themethod for compounds 1-10.

As in our previous method,16 we used control chartsas measure for a day-to-day system suitability check,which were constructed for 10 compounds chosen acrossthe entire range and which are indicated in Table 1 (seeStatistical Analysis section). An unexpected variationin these plots would immediately “flag” questionableresults. Furthermore, we find that triflupromazine (CASno. 146-54-3) offers a very sensitive “probe” for thecolumn performance monitoring, as well as its analoguechlorpromazine (CAS no. 50-53-3). A decline in the logk′w value of either of these two compounds is a goodindication of column deterioration. A log k′w value fortriflupromazine below 2.7 (2.6 for chlorpromazine)would indicate that the column should be replaced.

Currently, the ElogD determinations are run with theaqueous portion of the mobile phase prepared by acommercial laboratory, according to our specifications,and in fairly large batches (>20 gallons), and we havenoticed no significant difference in performance, afteran initial adjustment of pH as needed. This practicalenhancement helps with the speed of analysis, and itmay also be taken as an indication of the ruggedness ofthe method.

As a further improvement, in terms of speed ofanalysis, we have also attempted to increase the flowrate, and we have further automated the calculationprocedure through in-house software to obtain the finalElogDoct value, with very limited manual intervention,directly from the chromatographic data file. Thesemodifications allow for an enhanced throughput, start-ing with an already rapid procedure. ElogDoct data forany compound are obtained, on average, in e20 min,on a single instrument. Equation 3 shows a highcorrelation between the log k′w values obtained withthe “standard” flow rate (sf, 1.0 mL/min) and the fasterflow rates (ff, 1.5 mL/min), used for a “mixed” set of datacomprising 56 proprietary and commercial compounds,using the same compounds under each condition, andencompassing roughly 3 log k′w units, largely across themedium range defined in the Experimental Section. Wehave not yet implemented this flow rate in our routinework, while we have run over 2000 compounds with the“standard” flow rates.

Similarly, we have obtained good results by increasingthe flow rate of the high lipophilicity range, from 2 to 3mL/min (data not shown).

As a further caveat with the use of shake methodswe report the widely different results we obtained withguanoxan (a guanidine derivative, CAS no. 5714-04-5)for which a value of -0.83 was reported as logDoct.22

Using the shake-vial procedure B (see ExperimentalSection), values of -0.1 and -0.3 were obtained, induplicate determinations. In a fairly extensive logDoctvs concentration study, using the shake-vial procedureA, and we found that the values ranged from -1.6 to-1.0 upon decreasing the concentration, from 1.5 to 0.1mg, in a 50:2 octanol:buffer system. Indeed Perlman23

Figure 2. Plot of residuals vs logDoct.

Table 2. Extrapolated Capacity Factors and logDoct Data for10 Proprietary Compounds

compd log k′wa sdb ElogDoctc logDoct

d residualse

1 0.15 0.07 0.38 0.31 -0.072 0.79 0.03 1.10 1.16 0.063 2.75 0.25 3.31 3.20 -0.114 0.30 0.06 0.55 0.68 0.135 1.14 0.12 1.49 1.66 0.176 2.17 0.10 2.65 2.19 -0.467 3.46 0.02 4.11 3.37 -0.748 2.70 0.01 3.25 2.58 -0.679 3.68 0.16 4.35 3.85 -0.5010 2.14 0.06 2.62 2.10 -0.52

a Average of three to five determinations. b Standard deviationfor log k′w determinations. c Predicted ElogDoct from eq 1. d Datafrom shake-vial and/or potentiometric determinations. The averagewas taken if more than one value was available. e Averageresidual, - 0.27. ElogDoct vs logDoct: R2 ) 0.969.

log k′w(P) ) 1.0429 (( 0.0241)log k′w(D) -0.0219 (( 0.0522) (2)

N ) 36, R2 ) 0.982, R ) 0.991, s ) 0.198,F ) 1866, q2 ) 0.980

log k′w(sf) ) 0.8823 (( 0.0378)log k′w(ff) +0.1474 (( 0.0654) (3)

N ) 56, R2 ) 0.910, R ) 0.954, s ) 0.175,F ) 544, q2 ) 0.9

2494 Journal of Medicinal Chemistry, 2001, Vol. 44, No. 15 Lombardo et al.

has reported a large variation in the logDoct values ofdiarylguanidines, up to 2 logDoct units depending uponthe counterion present, and that might be the case here.Under our conditions we found an ElogDoct value of-0.3, which is in very close agreement with the datafrom shake-flask procedure B. However, it deviatedsignificantly from the values from procedure A, even atthe lowest concentration we have reached, and mightbe borderline acceptable for estimation, against theliterature data.

The present chromatographic method is not usable,at the moment, for the determination of the ElogDoct ofacidic compounds (significantly ionized at pH 7.4).However, the method offers high accuracy for com-pounds which are devoid of any significant ionizationand thus are not amenable to a logDoct determinationvia well-known potentiometric techniques.24 It alsooffers high accuracy, together with the other charac-teristics of the RP-HPLC method, for compounds havinglow solubility in water, without recourse to mixedsolvents and extrapolation, as in shake-flask and po-tentiometric techniques.

Conclusion

We have demonstrated that, by a judicious choice ofmobile phase and RP-HPLC column, a very accurateElogDoct determination method, for basic and neutralcompounds, could be developed. This method respondsto the criteria of rapid throughput, ruggedness, andminimal manual intervention set forth in the Introduc-tion, for drug-like compounds. We have also offeredsome practical considerations to help its application tohigh-throughput screening. Since logDoct has been shownto be an important parameter for the ADME, QSAR,and QSPR profiling of newly synthesized compounds,3,5

we believe this method will find useful applications inpharmaceutical discovery and development settings. Asa final comment we note that, with a minor adjustmentof pH, the method should be amenable of use for thedetermination of ElogDoct at pH 6.5, thus finding ap-plication in correlations specifically involving intestinalabsorption.5

Experimental Section

Materials and Methods. Most the solutes were purchaseddirectly from commercial sources (Aldrich, Fluka, ICN, RBI,Sigma, Tocris) and used as received, in all cases. In severalcases they were available through our Materials Managementgroup as either proprietary compounds or samples extractedfrom commercial formulations. Deionized water and HPLCgrade methanol (J. P. Baker) and 1-octanol (Fisher Scientific)were used throughout.

The mobile phase consisted, in all cases, of 20 mM MOPSbuffer at pH 7.4, with the addition of 0.15% of n-decylamine12,13

and methanol in varying proportions from 70 to 15% v/v. A0.25% (v/v) amount of octanol was added to methanol, andoctanol-saturated water was used to prepare the buffer. Themobile phase is now routinely prepared, according to ourspecifications, in larger batches, by Brand-Nu Labs, Meriden,CT.

The capacity factors data (k′ ) (tR - t0)/t0), obtained atvarious amounts of methanol, were then extrapolated to 0%methanol and reported as log k′w, using a linear procedure. Inall cases, except for allopurinol (R2 ) 0.96), the square of thecorrelation coefficient was 0.99. Injections of pure methanolwere used to determine t0, i.e., the dead time, while tR has theusual meaning of the retention time for the analyte. For very

low logDoct compounds, atenolol (CAS no. 29122-68-7) was usedto determine t0 and is now used routinely for the low range.

All the chromatographic runs were performed on an Agilent1100 HPLC ChemStation at the ambient temperature. TheHPLC column used was Supelcosil LC-ABZ, 5 µm, 4.6 × 50mm. A diode array detector was used to monitor signals at235, 255, 265, 275, and 310 nm. We also tested columnsmanufactured from different silica bond lots to ensure repro-ducibility. Samples were dissolved in 1:1 methanol/water in aconcentration range of 50-100 µg/mL. The flow rate was 0.5,1, or 2 mL/min, depending on the lipophilicity range. Threeexperimental lipophilicity ranges were established using, inall cases, three points for the extrapolation to k′w, as de-scribed in the table below.

The samples are placed in the appropriate range by esti-mating their lipophilicity via computed values or by priorexperience with a given class. Experimental values obtainedfrom runs outside the appropriate range are typically run denovo. However, a “screen” using a single injection at 75%methanol can be performed to “weed out” high lipophilicitycompounds. If the retention time is >1.1 min, at a flow rate of2 mL/min, the compound would likely yield an ElogDoct > 5.The user may then decide to adjust the conditions for thatcompound, such as the duration of each run, or to use such anestimated value, thus increasing the throughput and guardingagainst potential carry-overs.

In each case the entire group of samples is run before thecolumn is equilibrated to the next condition, in an automatedfashion. We typically start from the high range (highestmethanol content) and run, sequentially, all the compounds.For the low range it was found that a period of equilibrationbetween 1.5 and 2 h is needed. At the end of a complete runthe column is flushed with acetonitrile, at 2.0 mL/min, for 10-20 min.

The data analysis is then performed via an automatedprocedure relying on in-house software, which yields theElogDoct values (see Results and Discussion section), directlyfrom the chromatographic data files.

The shake-flask logDoct data, and in some cases data fromcountercurrent chromatography, were taken from the litera-ture, after careful evaluation of the experimental method andtemperature reported (generally between 20 and 25 °C) in theoriginal references or they were determined in-house. In somecases, data were not available or could not be determinedexperimentally due to the high lipophilicity of the compound.In such cases (clotrimazole and tolnaftate), a computed valuewas used. The shake-vial experimental measurements per-formed in-house (Procedure A) were all conducted at least induplicate, in amber glass vials, and in some cases, with varyingratios of octanol and MOPS buffer, always mutually presatu-rated prior to the experiment. The vials were shaken at leastovernight. HPLC analysis at different wavelengths, aftercentrifugation and separation of the phases, was used for thequantitative analysis, and both phases were analyzed. In somecases compounds were sent to PGRD, Sandwich Laboratories,Sandwich, U.K., for a semiautomated logDoct shake-vial de-termination, using a phosphate buffer at pH 7.4 as the aqueousphase. In this case (Procedure B), a 1:1 ratio of n-octanol andbuffer (both phases were mutually presaturated) was used,generally with an agitation time g30 min, followed by cen-trifugation and analysis of both phases.

In several instances, as indicated in Table 1, the data wereobtained from pH-metric determinations performed by pIonInc., Woburn, MA, after submission of commercial or propri-etary samples.

Statistical Analysis. All regression analyses were per-formed via the JMP statistical software package (v. 3.2.1, SAS

ElogDoct range flow rate (mL/min) % MeOH

<1 0.5 15, 20, 251-3 1 40, 45, 50>3 2 60, 65, 70

RP-HPLC Lipophilicity Determination Journal of Medicinal Chemistry, 2001, Vol. 44, No. 15 2495

Institute Inc.). Ten compounds were selected across the set of90 compounds, covering the entire range of lipophilicity, tomonitor the day-to-day performance of the method. Statisticalcalculations showed that the use of the 10 compounds wouldensure that the estimated slope, in the final regressionequation, would be within (0.09 of true one. The JMP softwarewas also used for the quality monitoring. Data accumulatedfor the standard set of compounds and regularly plotted onthe control charts constitute a powerful method for thedetection of trends and variations in performance. Variationsin log k′w values, for the selected compounds, should notexceed (3ks, where ks is the standard deviation estimate basedon data collected under well controlled experiments.16 Triflu-promazine log k′w is recorded for each run to ensure that thecolumn is performing suitably.

Acknowledgment. The authors thank Mr. Brian D.Bissett, for writing the data analysis software, and Drs.Alex Avdeef and Cynthia Berger (pIon, Inc.) for thepotentiometric determination of some of the logDoct dataprovided. The help of Mr. Anthony Harrison and Mr.Ken Anto-Awuakye (PGRD, Sandwich Laboratories,UK) in determining some of the logDoct values via theirshake-vial procedure is gratefully acknowledged.

References(1) Yalkowski, S. H.; Valvani, S. C. Solubility and Partitioning I:

Solubility of Nonelectrolytes in Water. J. Pharm. Sci. 1980, 69,912-922.

(2) Smith, D. A.; Jones, B. C.; Walker, D. K. Design of DrugsInvolving the Concepts and Theories of Drug Metabolism andPharmacokinetics. Med. Res. Rev. 1996, 16, 243-266.

(3) Scherrer, R. A.; Howard, S. M. Use of Distribution Coefficientsin Quantitative Structure-Activity Relationships. J. Med. Chem.1977, 20, 53-58.

(4) Smith, D. A. Physicochemical Properties in Drug Metabolismand Pharmacokinetics. In Computer-Assisted Lead Finding andOptimization; van de Waterbeemd, H., Testa, B., Folkers, G.,Eds.; VCHA, Weiley-VCH: Weinheim, 1997; Chapter 17, pp265-276.

(5) Yoshida, F.; Topliss, J. G. QSAR Model for Drug HumanBioavailability. J. Med. Chem. 2000, 43, 2575-2585.

(6) Avdeef, A. Assessment of Distribution-pH Profiles. In Lipophi-licity in Drug Action and Toxicology; Pliska, V., Testa, B., vande Waterbeemd, H., Eds.; VCH: Weinheim, 1996; Chapter 5,pp 73-87.

(7) ACD/LogD Suite, v. 4.5. Advanced Chemistry Development:Toronto, Ontario, Canada.

(8) Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J.Experimental and computational approaches to estimate solubil-ity and permeability in drug discovery and development. Adv.Drug. Del. Rev. 1997, 23, 3-25.

(9) Sangster, J. Octanol-Water Partition Coefficients: Fundamen-tals and Physical Chemistry; Wiley: New York, 1997; pp 79-112.

(10) van de Waterbeemd, H. M.; Kansy, M.; Wagner, B.; Fischer, H.Lipophilicity Measurements by Reversed-Phase High Perfor-mance Liquid Chromatography (RP-HPLC). In Lipophilicity inDrug Action and Toxicology; Pliska, V., Testa, B., van deWaterbeemd, H., Eds.; VCH: Weinheim, 1996; Chapter 5, pp73-87.

(11) Pagliara, A.; Khamis, E.; Trinh, A.; Carrupt, P.-A.; Tsai, R.-S.;Testa, B. Structural Properties Governing Retention Mecha-nisms on RP-HPLC Stationary Phases used for LipophilicityMeasurements. J. Liq. Chromatogr. 1995, 18, 1721-1745.

(12) Minick, D. J.; Frenz, J. H.; Patrick, M. A.; Brent, D. A. AComprehensive Method for Determining Hydrophobicity Con-stants by Reversed-Phase High-Performance Liquid Chroma-tography. J. Med. Chem. 1988, 31, 1923-1933.

(13) Morelock, M. M.; Choi, L. L.; Bell, G. L.; Wright, J. L. Estimationand Correlation of Drug Water Solubility with PharmacologicalParameters Required for Biological Activity. J. Pharm. Sci. 1994,83, 948-952.

(14) Abraham, M. H.; Chadha, H. S.; Leo, A. J. Hydrogen Bonding.35. Relationship between high-performance liquid chromatog-raphy capacity factors and water-octanol partition coefficients.J. Chromatogr. A 1994, 685, 203-211.

(15) Yamagami, C.; Araki, K.; Ohnishi, K.; Hanasato, K.; Inaba, H.;Aono, M.; Ohta, A. Measurement and Prediction of Hydrophobic-ity Parameters for Highly Lipophilic Compounds: Applicationof the HPLC Column-Switching Technique to Measurement oflogP of Diarylpyrazines. J. Pharm. Sci. 1999, 88, 1299-1304.

(16) Lombardo, F.; Shalaeva, M. Y.; Tupper, K. A.; Gao, F.; Abraham,M. H. ElogPoct: A Tool for Lipophilicity Determination in DrugDiscovery. J. Med. Chem. 2000, 43, 2922-2928.

(17) Minick, D. J.; Brent, D. A.; Frenz, J. Modeling Octanol-WaterPartition Coefficients by Reversed-Phase Liquid Chromatogra-phy. J. Chromatogr. 1989, 461, 177-191.

(18) Unger, S. H.; Chiang, G. H. Octanol-physiological buffer distri-bution coefficients of lipophilic amines by reversed-phase high-performance liquid chromatography and their correlation withbiological activity. J. Med. Chem. 1981, 24, 262-270.

(19) The LC-ABZ column is considered to be “RP-18 like”, and it wasdescribed as being electrostatically coated to decrease the soluteinteractions with free silanols. It is described by the manufac-turer as an embedded amide functional group column and ashaving some polymeric (cross-linking) character and multiplepoints of attachment to the silica surface.

(20) Melander, W.; Stoveken, J.; Horvath, C. Stationary Phase Effectsin Reversed-Phase Chromatography. I. Comparison of Energeticsof Retention on Alkyl-Silica Bonded Phases. J. Chromatogr.1980, 199, 35-56.

(21) Valko, K.; Bevan, C.; Reynolds, D. Chromatographic Hydropho-bicity Index by Fast-Gradient RP-HPLC: A High-ThroughputAlternative to logP/logD. Anal. Chem. 1997, 69, 2022-2029.

(22) Camenisch, G.; Alsenz, J.; van de Waterbeemd, H.; Folkers, G.Estimation of permeability by passive diffusion through Caco-2cell monolayers using the drugs’ lipophilicity and molecularweight. Eur. J. Pharm. Sci. 1998, 6, 313-319.

(23) (a) Perlman, M. E.; Zhang, L.; Hu, L.-Y.; Reddy, N. L.; Fischer,J. B.; Durant, G. J. Use of pH dependence of logD for pKadetermination and formulation of diarylguanidines. Pharm. Res.1996, 13, S-33. (b) Pauletti, G. M.; Perlman, M. E.; Zhang, L.;Borchardt, R. T. Can counterions affect the transport of alipophilic diarylguanidine (CNS1237) across Caco-2 cell mono-layers in vitro? Pharm. Res. 1997, 14, S-24. We thank Dr.Perlman for a copy of both posters. (c) Maillard, M. C.; Perlman,M. E.; Amitay, O.; Baxter, D.; Berlove, D.; Connaughton, S.;Fischer, J. B.; Guo, J. Q.; Hu, L.-Y.; McBurney, R. N.; Nagy, P.I.; Subbarao, K.; Yost, E. A.; Zhang, L.; Durant, G. J. Design,Synthesis, and Pharmacological Evaluation of ConformationallyConstrained Analogues of N,N′-Diaryl- and N-Aryl-N-aralky-lguanidines as Potent Inhibitors of Neuronal Na+ Channels. J.Med. Chem. 1998, 41, 3048-3061.

(24) Slater, B.; McCormack, A.; Avdeef, A.; Comer, J. E. A. pH-Metriclog P. 4. Comparison of Partition Coefficients Determined byHPLC and Potentiometric Methods to Literature Values. J.Pharm. Sci. 1994, 83, 1280-1283.

(25) Barbato, F.; Caliendo, G.; LaRotonda, M. I.; Morrica, P.; Silipo,C.; Vittoria, A. Relationships between octanol-water partitiondata, chromatographic indices and their dependence on pH in aset of beta-adrenoceptor blocking agents. Farmaco 1990, 45,647-663.

(26) Stopher, D.; McClean, S. An improved method for the determi-nation of distribution coefficients. J. Pharm. Pharmacol. 1990,42, 144.

(27) LaRotonda, M. I.; Amato, G.; Barbato, F.; Silipo, C.; Vittoria, A.Relationship between Octanol-Water partition Data, Chromato-graphic Indices and their Dependence on pH in a Set ofNonsteroidal Antiinflammatory Drugs. Quant. Struct. Act. Relat.1983, 2, 168-173.

(28) Fujita, T.; Iwasa, J.; Hansh, C. A new substituent constant, π,derived from partition coefficients. J. Am. Chem. Soc. 1965, 86,5175-5180.

(29) Bundgaard, H.; Falch, E. Allopurinol prodrugs. II. Synthesis,hydrolysis kinetics and physicochemical properties of variousN-acyloxylmethyl allopurinol derivatives. Int. J. Pharm. 1985,24, 307-325.

(30) Carelli, V.; Di Colo, G.; Nannipieri, E.; Serafini, M. F. Enhance-ment effects in the permeation of Alprazolam through hairlessmouse skin. Int. J. Pharm. 1992, 88, 89-97.

(31) Sirius Technical Application Notes, 1995, vol. 2. Sirius AnalyticalInstruments, Ltd.: Forest Row: East Sussex RH18 5DW, U.K.

(32) Barbato, F.; LaRotonda, M. I.; Quaglia, F. Chromatographicindices determined on an immobilized artificial membrane (IAM)column as descriptors of lipophilic and polar interactions of4-phenyldihydropyridine calcium-channel blockers with biomem-branes. Eur. J. Med. Chem. 1996, 31, 311-318.

(33) Okada, J.; Esaki, T.; Fujieda K. 13C nuclear magnetic resonancespectra of antipyrine derivatives and their application to Hanschanalysis. Chem. Pharm. Bull. 1976, 24, 61-71.

(34) Lullmann, H.; Wehling, M. The binding of the drugs to differentpolar lipids in vitro. Biochem. Pharmacol. 1979, 28, 3409-3415.

(35) Hansch, C.; Leo, A.; Hoekman, D. Exploring QSAR. Hydrophobic,Electronic, and Steric Constants; American Chemical Society:Washington, DC; 1995; pp 3-216.

(36) Lacko, L.; Wittke, B. The affinities of benzodiazepines to thetransport protein of glucose in human erythrocytes. Arzneim.-Forsch. 1984, 34, 403-407.

2496 Journal of Medicinal Chemistry, 2001, Vol. 44, No. 15 Lombardo et al.

(37) Muller, W.; Wollert, U. Interaction of Benzodiazepines withHuman Serum Albumin. Circular Dichroism Study. Arch. Phar-macol. 1973, 278, 301-312.

(38) Taylor, P. J. Hydrophobic Properties of Drugs. In ComprehensiveMedicinal Chemistry; Hansch, C., Sammes, P. G., Taylor, J. B.,Eds.; Pergamon Press: Oxford, 1990; Vol. 5, pp 241-294.

(39) Henczi, M.; Nagy, J.; Weaver, D. F. Determination of octanol-water partition coefficients by an HPLC method for anticonvul-sant structure-activity studies. J. Pharm. Pharmacol. 1995, 47,345-347.

(40) Ter Laak, A. M.; Tsai, R. S.; Donne-Op den Kelder, G. M.;Carrupt, P.-A.; Testa, B.; Timmerman, H. Lipophilicity andhydrogen-bonding capacity of H1-antihistaminic agents in rela-tion to their central sedative side-effects. Eur. J. Pharm. Sci.1994, 2, 373-384.

(41) Sirius Technical Application Notes, 1994, vol. 1. Sirius AnalyticalInstruments, Ltd.: Forest Row: East Sussex RH18 5DW, U.K.

(42) Berthod, A.; Carda-Broch, S.; Garcia-Alvarez-Coque, M. C.Hydrophobicity of ionizable compounds. A theoretical study andmeasurements of diuretic octanol-water partition coefficientsby countercurrent chromatography. Anal. Chem. 1999, 71, 879-888.

(43) Timmermans, P. B. M. W. M.; Brands, A.; Van Zwieten, P. A.Lipophilicity and brain disposition of clonidine and structurallyrelated imidazolidines. Naunyn-Schmiedeberg’s Arch. Pharma-col. 1977, 300, 217-226.

(44) Carmichael, F. J.; Israel, Y. In vitro inhibitory effects of narcoticanalgesics and other psychotropic drugs on the active uptake ofnorepinephrine in mouse brain tissue. J. Pharmacol. Exp. Ther.1973, 186, 253-260.

(45) Kaufman, J. J.; Semo, N. M.; Koski, W. S. Microelectrometrictitration measurement of the pKa’s and partition and drugdistribution coefficients of narcotic antagonists and their pH andtemperature dependence. J. Med. Chem. 1975, 18, 647-655.

(46) Winiwarter, S.; Bonham, N. M.; Ax, F.; Hallberg, A.; Lennernas,H.; Karlen, A. Correlation of human jejunal permeability (invivo) of drugs with experimentally and theoretically derivedparameters. A multivariate data analysis approach. J. Med.Chem. 1998, 41, 4939-4949.

(47) Tomida, H.; Yotsuynagi, T.; Ikeda, K. Solubilization of steroidhormones by polyoxyethylene lauryl ether. Chem. Pharm. Bull.1978, 26, 2832-2837.

(48) Seiler, P.; Zimmermann, I. 5-phenyl-1,3-dihydro-1,4-benzodiaz-epin-2-ones. Arzneim.-Forsch. 1983, 33, 1519-1522.

(49) Meuldermans, W. E. G.; Hurkmans, R. M. A.; Heykants, J. J.P. Plasma protein binding and distribution of fentanyl, sulfen-tanil, alfentanil and lofentanil in blood. Arch. Int. Pharmacodyn.1982, 257, 4-19.

(50) Alessi-Severini, S.; Coutts, R. T.; Jamali, F.; Pasutto, F. M.Flecainide. In Analytical Profiles of Drug Substances andExcipients; Brittain, H. G., Ed.; Academic Press: San Diego,1992; Vol. 21, pp 169-195.

(51) Jezequel, S. G. Fluconazole: interspecies scaling and allometricrelationships of pharmacokinetic properties. J. Pharm. Phar-macol. 1994, 46, 196-199.

(52) Graf, E.; El-Menshawy, M. pK- and Vk-messungen an benzodi-azepinen. Pharm. Uns. Zeit. 1977, 6, 171-178.

(53) Schutz, V. H.; Fitz, H. Screening, detection and biotransforma-tion of Lormetazepam, a new hypnotic agent from the 1,4-benzodiazepine series. Arzneim.-Forsch. 1982, 32, 177-183.

(54) Anderson, R. F.; Patel, K. B. Effect of lipophilicity of nitroami-dazoles on radiosensitization of hypoxic bacterial cells in vitro.Br. J. Cancer 1979, 39, 705-710.

(55) Adams, G. E.; Clarke, E. D.; Flockhart, I. R.; Jacobs, R. S.; Sehmi,D. S.; Stratford, I. J.; Wardman, P.; Watts, M. E. Structure-activity relationships in the development of hypoxic cell radi-osensitizers. I. Sensitization efficiency. Int. J. Radiat. Biol. 1979,35, 2, 133-150.

(56) Wozniak, T. J. Nizatidine. In Analytical Profiles of DrugSubstances; Florey, K., Ed.; Academic Press: San Diego, 1990;Vol. 19, pp 397-427.

(57) Ungell, A.-L.; Nylander, S.; Bergstrand, S.; Sjoberg, Å.; Lenner-nas, H. Membrane transport of drugs in different regions ofintestinal tract of the rat. J. Pharm. Sci. 1998, 87, 360-366.

(58) Mohler, V. W.; Soder, A. On chemistry and synthesis of 3,7-dimethyl-1-(5-oxo-hexyl)-xanthine. Arzneim.-Forsch. 1971, 21,1159-1160.

(59) Buchi, V. J.; Oey, L. T.; Perlia, X. Zusammenhange zwischender chemischen Konstitution, den physikalisch-chemischen Eigen-schaften, der chemischen Reaktivitat und der lokalanasthetis-chen Wirksamkeit einiger Procain-Isostere. Arzneim.-Forsch.1972, 22, 1071-1084.

(60) Manners, C. N.; Payling, D. W.; Smith, D, A. Lipophilicity ofzwitterionic sulphate conjugates of tiaramide, propranolol and4′-hydroxypropranolol. Xenobiotica 1989, 19, 1387-1397.

(61) Tsai, R.-S.; Carrupt, P.-A.; Testa, B.; Tayar, N. E.; Grunewald,G. L.; Casy, A. F. Influence of Stereochemical Factors on thePartition Coefficient of Diastereomers in a Biphasic Octan-1-ol/water System. J. Chem. Res. (M), 1993, 1901-1920.

(62) Iwasa, J.; Fujita, T.; Hansch, C. Substituent constants foraliphatic functions obtained from partition coefficients. J. Med.Chem. 1965, 8, 150-153.

(63) O’Connor, D. O.; Capel, C.; Rycroft, W.; Tattersall, F. D.; Locker,K., Sohal, B.; Graham, M. I.; Evans, D. C. Influence of thePhysicochemistry on the Brain Penetration of the ‘Triptans inRat. Poster presented at the XIV Course in Drug Research, June5-6, 1997, Helsinki, Finland.

(64) Artursson, P.; Karlsson, J. Correlation between Oral DrugAbsorption in Humans and Apparent Drug permeability Coef-ficients in Human Intestinal Epithelial (CACO-2) Cells. Biochem.Biophys. Res. Comm. 1991, 175, 880-885.

(65) Hansch, C.; Nakamoto, K.; Gorin, M.; Denisevich, P.; Garrett,E. R.; Heman-Ackah, S. M.; Won, C. H. Structure-activityrelationship of chloramphenicols. J. Med. Chem. 1973, 16, 917-922.

(66) Barbato, F.; Recanatini, M.; Silipo C.; Vittoria, A.; Hydrophobicconstants and quantitative structure activity relationships(QSAR) in sets of phenothiazine drugs. Eur. J. Med. Chem. 1982,17, 229-234.

(67) Priewer, H.; Ullrich, F.; Herzfeldt, C. D.; Kleinsorge, D.; Mut-schler, E. Solubility and distribution coefficients of potassiumand magnesium retaining triamterene derivatives. Pharmazie1996, 51, 279-282.

JM0100990

RP-HPLC Lipophilicity Determination Journal of Medicinal Chemistry, 2001, Vol. 44, No. 15 2497